Scribd
Upload
26 views

Lecture 3 - AI and Learning For PDF

This document discusses deriving value from data through data mining and artificial intelligence techniques. It provides an overview of key concepts like data science, data mining, machine learning, deep learning and their applications. It also discusses limitations of early machine learning and how advances like neural networks, deep learning and new hardware are helping AI replicate some human abilities like rapid learning.

Uploaded by

Shabeer Ali PC
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
26 views

Lecture 3 - AI and Learning For PDF

This document discusses deriving value from data through data mining and artificial intelligence techniques. It provides an overview of key concepts like data science, data mining, machine learning, deep learning and their applications. It also discusses limitations of early machine learning and how advances like neural networks, deep learning and new hardware are helping AI replicate some human abilities like rapid learning.

Uploaded by

Shabeer Ali PC
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 24

AcSIR Course : Computer Applications and

Informatics

Deriving Value from Data by Data Mining and


Artificial Intelligence techniques

October 2021
Elizabeth Jacob
CSIR-NIIST
Summary so far
1. Cloud Technology offers unlimited resources of
memory, storage and computing on a pay-as-you use
model.
2. Big Data is generated by humans and machines – the
5 dimensions of Big data are Volume, Variety,
Velocity and Veracity and Value.
3. Sensor data from IoT is a producer of Big Data.
4. Big Data requires Analytics for deriving useful
information. – data mining and AI
5. Big Data resides in the Cloud and the Analytics
Software Platforms are also available on cloud to buy.
 Data Science
 Data Mining
 AI - Machine Learning and Deep Learning
 Future of AI
Data Science
• Data science is an interdisciplinary field focused
on extracting knowledge from data sets (small
and big) and applying the knowledge to make
actionable insights.
• The foundations of data science rest on
statistics, informatics, computer
science, machine learning, and development of
new technologies to gain insights from data.
Drowning in Data but Starving for Knowledge
John Naisbitt 1982
• Wide availability of huge amounts of data from
terabytes(10004) to yottabytes(10008)coming at
high velocities, knowledge discovery needed to
make sense and use of data.
• Data mining is the automatic extraction of
patterns, non-trivial insights, predictions,rules
and regularities from data in large repositories.
Symbolic of Mining the earth for precious
minerals
Example of Data Mining by Transaction Analysis
Market-Basket Analysis
The rule {bread} ->{butter, jam} found in the sales data
of a supermarket would indicate that if a customer
buys bread, they are likely to also buy butter or jam.
Time series clustering to discover
commonly purchased items that are useful
for formulating sales strategies.
Moravec’s paradox of AI
• In the 1980s, Hans Moravec, Rodney Brooks, Marvin
Minsky and others articulated this AI paradox

What is difficult for human is easy for the computer


BUT
What is easy for human is difficult for the computer
Some Machine Learning Techniques
• Regression - algorithms and techniques to find the
best fit to the underlying data.
• Classification
• Clustering
• Anomaly
Detection
for outliers

ML algorithms : Random Forest, Support Vector


Machine and Kernel Regression
Neural Networks

• A biologically inspired sub-field of artificial


intelligence modeled after the brain
conceptualized in 1980s.
"Artificial Neural Network" ANN
• The fundamental computing unit is the neuron
which takes multiple inputs and outputs an
activation function of the weighted sum of the
inputs.
How ANNs work

1. Information is fed into the input layer which


transfers it to the hidden layer
2. The interconnections between the two layers
assign weights to each input randomly
3. A bias added to every input after weights are
multiplied with them individually
4. The weighted sum is transferred to the
activation function
Between the input layer IL and output layer OL,
there are one or more hidden layers HL.
Contd. Backpropagation ANN Algorithm
5. The activation function determines which nodes it
should fire for feature extraction
6. The model applies an application function to the output
layer to deliver the output
7. Weights are adjusted, and the output is back-
propagated to minimize error
8. The model uses a cost function to reduce the error rate.
9. The model compares the output with the original result
10. The model adjusts the weights in every iteration to
enhance the accuracy of the output.
Limitations of Machine Learning

• Simple neural networks with 100s or even


1000s of neurons, connected in a relatively
simple manner, just couldn’t duplicate what
the human brain could do.
• Human brains have around 86 billion neurons
and very complex interconnectivity.
• Machine learning still got stuck in learning.
Deep Learning
• Deep learning is a family of techniques in AI and ML, a re-
discovery of neural networks with more neurons, layers,
and inter-connectivity.
• Each algorithm in the hierarchy applies a nonlinear
transformation to its input and uses what it learns to
create a statistical model as output. Iterations continue
until the output has reached an acceptable level of
accuracy.
• The number of processing layers or hidden layers through
which data must pass is what inspired the label deep.
• Big data and more processing power make the ANN grow
deeper to solve more difficult problems.
Application of AI
ElemNet: Deep Learning the Chemistry of
Materials From Only Elemental Composition

Sci Rep 8, 17593 (2018)


Application of AI – Speeding up the
materials discovery process
• If climate change proceeds as expected, humanity
might have a couple of decades only to come up with
materials that don’t yet exist: molecules that enable
us to quickly and cheaply capture carbon, and
batteries—made of something other than lithium, a
metal that is costly and difficult to mine—to store
the global supply of renewable energy- Alán Aspuru-
Guzik
• Aspuru-Guzik’s is building a lab in Toronto where AI
algorithms design novel molecules, and robots
quickly make and test them.
How AI is reinventing Computers- Hardware

• Unlike traditional chips, which are geared toward


ultrafast, precise calculations, Google’s tensor
processing units TPUs are designed for high-
volume but low-precision calculations required by
neural networks.
• Google has used these chips in-house since 2015:
they process people’s photos and natural--
language search queries. Google’s sister company
DeepMind uses them to train its AI applications.
How AI is reinventing Computers -Software

• The second change concerns how computers are told


what to do. For the past 40 years we have been
programming computers; for the next 40 we will be
training them.
• Programmers no longer write rules. Instead, they
create a neural network that learns those rules for
itself.
• AlphaZero taught itself to play Go better than humans.
Similarly, AlphaFold cracked open a biology problem—
working out how proteins fold—that people had
struggled with for decades.
The new era of Computing
• Getting a computer to do something meant typing
in a command, or at least clicking a button.
• Machines no longer need a keyboard or screen for
humans to interact with- they can see and hear
(Alexa)
• Humans want to spend less time telling them
what to do. They should be able to work out what
we need without being told.
• This is the shift from number-crunching to
decision-making.
AI is slowly overcoming Moravec’s paradox.
As Higher-level AI is beginning to replicate our
evolutionary abilities.

You might also like